[OliverB]

Hi,

I have been following the forum for a long time and today here is my first post. I am a graduate in biology and am learning the cfd by myself.

My question is about modeling a stirred tank with pitched blade turbine (RAS with standard wall function).

I performed a grid sensitivity study on 3 meshes (200, 400 and 800 K cells) and found that the error increase on the last. When I checked Y+ I found low values (average 10 and minimum less than one).

I change SHM settings to get a coarser mesh (500K) and get acceptable Y+ values, but under 30 in upper part of the tank and in part of blades near hub where velocity is low (see attachments 500K and Y+).

In order to go further I performed a finer mesh (800 K) but without refinement of the stirrer to increase Y+ on it. But the snapping is now not satisfying (see attachment).

Then, I was wondering :

- If I have part with low Y+ in region of low velocity what is the impact on results accuracy? Is the 500K mesh good enough ?

- How to refine a mesh while keeping Y+ high enough for wall function treatment ?

Thanks for your help.

[OliverB]

Hello, I continued my research in the forum and it seems that it is difficult to get a y+ on the whole wall in complex geometries.

I am not familiar with the underlying physics and have difficulty evaluating the impact of a first layer in the buffer layer where velocities are low, on the overall flow, while there is a large difference in velocity in the tank. I am not interested in the boundary layer or heat transfer but only in the overall flow and mixing time.

In this blog https://www.computationalfluiddynami...oundary-layer/ it is recommended to use scalable / enhanced wall treatment in my case but it is not available in openFoam.

I will try to optimize my mesh and refinements and redo grid sensitivity studies.